DEI: Heterogeneous LLMs Enhance Quality-Diversity Search

There's this new distributed framework for Quality-Diversity search called DEI, which stands for Diversity in Evolutionary Inference. Unlike standard searches that stick to similar models, DEI takes advantage of different large language models (LLMs) as mutation operators between nodes. Each LLM brings its own creative flair, adding diversity to the behavior. It builds on something called the Digital Red Queen framework, where nodes share their best solutions after each round to kick off the next stage, creating competitive interactions among models. In tests using Core War, a competitive programming environment, a four-node mix of models—including GPT-5.4-mini and Claude Sonnet 4.6—performed really well.